TCR γδ cell–specific STAT5 gain of function induces a druggable chronic human immune dysregulation

Susceptibility to infection is orchestrated by the host’s immune system. Besides numerous secondary causes of immunodeficiency (1), inborn errors of immunity (IEI) are a rapidly growing group of immune dysregulation entities due to germline mutations in nonredundant immune system genes (2). In addition, somatic mutations in the same genes may lead to phenocopies of IEI (3). IEI are still not diagnosed in many patients, or diagnosis is delayed for many years (4).

Signal transducer and activator of transcription 5 (STAT5) proteins play a key role in regulating hematopoiesis and hepatocyte function (5). STAT5A and STAT5B transcription factors form homo- or heterodimeric complexes that regulate differential gene expression (5, 6). The STAT5 dimerization is initiated upon binding of cytokines such as interleukin-2 (IL-2) or growth factors to their respective receptors on the cell surface. This process triggers the signaling via JAK1, JAK2, and JAK3 tyrosine kinases, leading to the nuclear translocation and transcriptional activities of STAT5 (5, 7).

STAT5 insufficiency resulting from germline biallelic loss-of-function mutations or dominant-negative monoallelic mutations can cause eczema, autoimmunity, and growth hormone–insensitive growth failure (8, 9). A germline STAT5 gain-of-function (GOF) mutation has been identified in two patients with atopic disease (10). In addition, somatic GOF mutations causing STAT5 overactivity drive inflammation, autoimmunity, myeloid, or lymphoproliferative diseases, and promote cancer, depending on the mutated cell lineage (5, 11, 12, 13). In malignant cells, persistent activation of STAT5 facilitates evasion of cell cycle control, playing a crucial role in leukemogenesis and the proliferation of various tumors (14).

T lymphocytes can express TCRs that result from pairing either α and β chains (TCR αβ) or γ and δ chains (TCR γδ). In humans, TCR γδ cells constitute 0.5–16% of the total circulating T cells and are present at higher frequencies in peripheral tissues (15). The TCRs γδ can recognize a broad range of molecules, including various members of the butyrophilin family (16). Most circulating TCR γδ cells in humans express a TCR Vγ9Vδ2 heterodimer and are activated by phosphorylated metabolites shared among several bacteria (17). The widespread distribution of bacterial phosphoantigens (pAgs) leads to continuous activation and supports their expansion. This constant stimulation induces TCR Vγ9Vδ2 cells to mature into effector cells that are ready to respond upon the subsequent antigen encounter. Consequently, these T cells are considered intermediate protagonists of both innate and adaptive immunity (18).

Dysregulation of TCR γδ cells contributes to the pathogenesis of systemic sclerosis, psoriasis, rheumatoid arthritis, and systemic lupus erythematosus, among others (18, 19, 20). Upon activation, they more frequently display T helper 1 (Th1) cytokine profiles, but may also rarely exhibit Th2 or Th17 profiles, secreting either IFN-γ, TNF-α, and IL-2; IL-4 and IL-13; or IL-17, IL-21, and IL-23, respectively (21, 22, 23). These pleiotropic functions contribute to their participation in diverse immune responses.

Both TCR αβ and TCR γδ T cells may appear as T cell large granular lymphocytes (T-LGL) in peripheral blood, with an activated morphology on blood smear microscopy and flow cytometric expression of typical natural killer (NK) cell markers such as CD16, CD56, and/or CD57. T-LGL are not pathologic per se, and increased numbers (reactive expansions) of T-LGL are observed during infections or inflammation (24). In contrast, T-LGL leukemia is referred to as a clonal, distinct (typically >2 x 10⁹/L) expansion of T-LGL that may be complicated by cytopenias, lymphoproliferation, and/or autoimmune diseases (24).

Here, we describe several patients with chronic immune dysregulation, normal lymphocyte counts, but chronically elevated TCR γδ vs. TCR αβ ratios. In one of the patients, molecular analysis revealed a TCR γδ–specific STAT5 GOF mutation as the molecular driver. Tailored treatment with the JAK inhibitor baricitinib successfully normalized hyperinflammation and reduced peripheral TCR γδ cells.

We screened our prospective multicenter adult immune dysregulation cohort, currently comprising >600 patients. We identified five individuals with normal absolute lymphocyte counts who exhibited a reduced TCR αβ/TCR γδ ratio as the primary immunophenotypic abnormality (Table 1). All were adults with chronic susceptibility to airway infections, mainly chronic sinusitis. None of the patients displayed antibody deficiency. In all patients, the increased proportion of TCR γδ cells and the normal total lymphocyte counts remained stable over time (most had several years of follow-up) (Table 1). The TCR γδ cells were classified as T-LGL by the diagnostic hematology flow cytometric assessment, based on variable expression of the NK cell markers CD16, CD56, and CD57 (Table 1). Next-generation sequencing (NGS)–based TCR sequencing was performed in four patients, revealing that >85% of the TCR γδ cell expansion was polyclonal in all (Table 1). In the remaining patient, PCR-based testing showed biclonal TCR expansion over a polyclonal background (Table 1). The goal of this study was to characterize potential molecular culprits driving the predominant polyclonal relative increase in TCR γδ cells in these patients and to study TCR γδ cell function.

P1 is a Caucasian male who has been monitored at our hospital for >16 years, starting at age 24. At his first visit, P1 presented with recurrent, large, and multiple oral erosive lesions that occurred twice monthly since young adulthood (Fig. 1 A), accompanied by chronic arthralgia and fatigue. P1 reported a susceptibility to prolonged, mostly mild (viral) upper respiratory tract infections that typically did not require antibiotic treatment. Family history for similar manifestations was negative. P1 repeatedly showed mild, inconsistent thrombocytopenia, monocytosis, and variable neutrophilia and eosinophilia (Fig. S1 A). Serum IgG and IgA were within the reference range, while IgE (456 IU/ml) and IgM (2.54 g/L, polyclonal) were slightly elevated (Fig. S1 B). Hemoglobin levels remained normal at all collection times, as was the C-reactive protein (except in acute infections, where it was mildly elevated). Repeated sonography and computed tomography (CT)–based imaging showed neither lymphadenopathy nor splenomegaly (Fig. S1 C). Additional immunologic abnormalities included increased positive antinuclear autoantibodies with a nuclear dot immunofluorescence staining pattern, and elevated CD21^low B cells (Fig. S1 B). The latter may indicate hyperinflammation driven by IFN-γ (25). The patient never met diagnostic criteria for systemic lupus erythematosus. Pathergy skin testing and HLA-B51 tests were negative, arguing against Behcet’s syndrome. A bone marrow biopsy revealed normal erythro- and myelopoiesis, along with reactive interstitial lymphocytic infiltration, without signs of lymphoma. Flow cytometry analysis of peripheral blood mononuclear cells (PBMCs) confirmed the overrepresentation of CD4⁻CD8⁻ T cells, comprising ∼30% of all CD3⁺ T cells (Fig. 1 B and Fig. S1 D). Most of these T cells expressed TCR γδ (Fig. 1 B). The relative increase in TCR γδ vs. TCR αβ cells remained stable over the following 9 years in the absence of immune-modulating therapy (Fig. 1, C and D). The frequencies of NK cells and the different TCR αβ cell subpopulations (naive, central memory, effector memory, CD4 regulatory T [Tregs], CD4 Tfh) were within normal ranges, except for Tregs, which were slightly relatively increased (Table 2). NGS-based TCRγ sequencing revealed seven different clones with >1% frequency, the top clone accounting for 6.7% (Table 1 and Table 3). Thus, the majority of the relative TCR γδ expansion was polyclonal.

To investigate a possible genetic predisposition for the preferential presence of TCR γδ cells, we performed whole-exome sequencing (WES) of P1’s flow cytometry–sorted αβ vs. γδ T cell subpopulations. We filtered the WES results for rare immune system gene variants with low allele frequency (<0.01) and high combined annotation-dependent depletion (CADD) score (>15), revealing a c.A1994T; p.Y665F STAT5B variant present in TCR γδ cells only (Fig. 1, E and F; and Table 4), in 14 out of 68 reads (21%). This suggested that the STAT5B variant was not restricted to one TCR clone, as the top TCR γδ clone frequency in P1 was 6.7% (Table 1 and Table 3). The identified STAT5B missense variant is located within the STAT5B SH2 domain (Fig. 1, G and H). This variant has been previously shown to cause STAT5 GOF, displaying enhanced DNA binding, while having a less significant impact on STAT5 phosphorylation (26). We confirmed that the c.A1994T mutation was expressed at the mRNA level through Sanger sequencing of a PCR product generated from cDNA of sorted TCR γδ cells (Fig. 1 F). The STAT5B mutation was not detected by Sanger sequencing in various enriched myeloid lineage populations (monocytes, granulocytes, and eosinophils) of P1 (Fig. S1, E–G).

WES analysis of gDNA from flow cytometry–sorted TCR γδ cells of one other patient (cohort number 86) detected neither STAT3 nor STAT5 mutations that could have explained the expansion of TCR γδ cells. In the remaining three other patients, we have as of yet only performed targeted Sanger sequencing of the corresponding STAT5B region using gDNA from bulk PBMCs, without the detection of a somatic mutation (Table 1). Bulk sequencing may, however, miss low-frequency TCR γδ cell–restricted somatic mutations.

To validate STAT5B activity in primary T cells of P1, we measured STAT5 phosphorylation in TCR αβ vs. TCR γδ cells isolated from PBMCs from P1 vs. five healthy controls (HCs) using flow cytometry (Fig. 1, I and J). Sorted TCR αβ and TCR γδ cells were rested overnight, and stimulated (or not) with recombinant IL-2 for 30 min. After IL-2 stimulation, P1-derived TCR γδ cells showed a significantly higher percentage of phospho-STAT5^high (pSTAT5) cells compared with HC, while TCR αβ cells showed no significant difference in STAT5 phosphorylation (Fig. 1 J). Similar results were observed for the measurements of pSTAT5 Median fluorescence intensity (MFI) (Fig. S1 H). In the absence of IL-2 stimulation, TCR αβ cells and TCR γδ cells showed no differences in pSTAT5 MFI (Fig. S1 I). These data indicate that the missense mutation drives STAT5 hyperphosphorylation in TCR γδ cells upon IL-2 stimulation.

T cells isolated from PBMCs of P1 and two HCs were further characterized using multicolor flow cytometry (gating strategy displayed in Fig. S2 A). This confirmed the reduced frequency of TCR αβ cells and an increased frequency of TCR γδ cells in P1, and demonstrated that the majority of TCR γδ cells were TCR Vδ2 cells (Fig. 2, A and B). We isolated TCR Vδ2 from P1-derived PBMCs and confirmed the presence of the mutated STAT5B in the TCR Vδ2 subset (Fig. S2 B). TCR Vδ2 cells of both P1 and HCs primarily displayed a CD4⁻CD8⁻ phenotype, representing ∼80%, while the remaining (∼20%) expressed only CD8 (Fig. 2, C and D). The differentiation of the TCR Vδ2 subset was assessed based on the expression of CD45RA and CCR7 (Fig. 2 E). TCR Vδ2 cells from both P1 and HCs were primarily T effector memory (TEM) or T effector memory re-expressing CD45RA (TEMRA), with frequencies of ∼65 and 30%, respectively (Fig. 2 F). 10 different surface markers were analyzed for the TCR Vδ2 population and were used in clustering analysis (Fig. S2 C; and Fig. 2, G and H).

The TCR Vδ2 cells from both P1 and HCs exhibited elevated KLRG1 levels, indicating that they were antigen-experienced (Fig. S2 C). The activation markers CD69 and CD26, the inhibitory molecule PD-1, the apoptosis-inducing mediator CD95, and the NK marker NKp80 were similarly expressed by TCR Vδ2 cells from both groups (Fig. S2 C). Instead, the NK marker CD56 and the costimulatory molecule CD28 were higher in TCR Vδ2 cells from P1 compared with HCs. Additionally, TIGIT was expressed in a subset of TCR Vδ2 cells from P1 (Fig. S2 C). Finally, CD57 was more abundantly expressed in TCR Vδ2 cells from HCs compared with P1, where it was nearly absent (Fig. S2 C). The expression of CD57 on lymphocytes may indicate reduced proliferative capacity (27), linking the CD57^low expression in P1-derived TCR Vδ2 cells to the high frequencies observed in the blood of P1 and to the STAT5B GOF mutation.

The clustering analysis of these markers revealed that clusters 10 (KLRG1⁻, CD28⁺, CD95⁺, PD-1⁺), 8 (KLRG1⁺, CD56⁺, CD57⁺), and 1 (KLRG1⁺) were more prevalent in HCs than in P1. Cluster 10 defines a population of PD-1⁺ TCR Vδ2 cells with high cytotoxic capacity linked to CD95 expression. However, the overall presence of this cluster was very low in both HCs and P1. Conversely, cluster 8, representing CD56⁺ TCR Vδ2 cells with low proliferation capacity related to CD57 expression, and cluster 1, which showed no expression of inhibitory or activation markers, constituted a substantial part of the TCR Vδ2 cells found in the periphery (Fig. 2, G and H). In summary, these phenotypic data suggest a decrease in nonactivated TCR Vδ2 cells, characterized by lower proliferative capacity, in P1. The enriched clusters identified in P1, including clusters 3 (KLRG1⁺, CD26^bright, CD56^bright), 9 (KLRG1⁺, CD26^bright, CD56^bright, NKp80⁺), and 11 (KLRG1⁺, CD26^bright, TIGIT⁺), indicate an enrichment of activated TCR Vδ2 cells with an NK-like phenotype (Fig. 2, G and H).

As a control, the clustering of the TCR Vδ2⁻ cells showed no significant differences in the measured markers. Cluster 14, which consists of nonactivated TCR γδ–negative cells, and cluster 5, featuring CD8⁺ and CD57⁺ TCR γδ–negative cells, were enriched in P1 and HCs, respectively. However, these clusters constitute only a small fraction of the TCR Vδ2–negative cells (Fig. S2, D and E).

STAT5 phosphorylation is essential for T cell proliferation (28). To investigate whether the STAT5B GOF mutation found in TCR Vγ9Vδ2 cells from P1 enhances their ability to proliferate in vitro, we stimulated PBMCs from three HCs and P1 with a suboptimal dose of the pAg (E)-4-hydroxy-3-methyl-but-2-enyl pyrophosphate (HMBPP) (0.4 nM). This low dose caused a distinct proliferation of TCR Vδ2 cells from P1, but not from HCs (Fig. 3 A). No other T cells were expanded from the HMBPP stimulation (Fig. S3, A and B). Despite differences in proliferative capacity under suboptimal HMBPP stimulation, TCR Vδ2 cells from both P1 and HCs exhibited similar proliferative potential when stimulated with PHA (Fig. 3 B), indicating that the response to TCR stimulation was enhanced in P1, while the reaction to other PHA-stimulated receptors was not. In all three HCs, suboptimal HMBPP doses did not alter the in vitro frequencies of TCR Vδ2 cells. However, in P1, it increased the percentage of TCR Vδ2 cells within CD3⁺ cells from 15 to 21% (Fig. 3 C). When activation marker expression was assessed, CD69, CD137, and ICOS were detected on most proliferating cells after just one cell division (Fig. 3 D and Fig. S3 C). CD39 was expressed in only about one third of proliferating cells and appeared after at least two divisions (Fig. 3 D and Fig. S3 C).

The activation of TCR Vγ9Vδ2 cell lines derived from P1 vs. HCs was further evaluated by measuring cytokine release. HMBPP stimulation caused a greater dose-dependent secretion of IFN-γ (Fig. 3 E) and TNF-α (Fig. 3 F) from HC lines compared with P1, indicating a preferential Th1 response in TCR Vγ9Vδ2 cell lines derived from HCs. In contrast, the P1-derived TCR Vγ9Vδ2 cell line shows a trend toward secreting more IL-4 (Fig. 3 G) and IL-13 (Fig. 3 H) than HC lines in response to 100 nM HMBPP. This suggests a distinct Th2-skewed cytokine profile in P1-derived TCR Vγ9Vδ2 cells, consistent with the clinical and laboratory phenotype.

To define the gene expression profile and differentiation program of the STAT5 GOF TCR γδ cells, we sorted TCR γδ cells from P1 and HCs and performed RNA sequencing (RNA-seq). The PCA showed that TCR γδ cells from P1 and HCs were transcriptionally distinct both in ex vivo and in in vitro overnight-rested TCR γδ cells (Fig. 4 A).

We also performed RNA-seq on TCR γδ cells from P1 and HCs, treated or not with IL-2 for 4 h before RNA extraction. PCA revealed that untreated and IL-2–treated TCR γδ cells clustered together for both P1 and HCs, indicating that IL-2 stimulation did not result in significant overall differential gene expression (Fig. 4 A).

Since STAT5 is essential for the IL-2 signaling pathway (29), we initially examined related genes (Fig. 4 B). Classical STAT5B target genes, such as IL-2RB, SOCS2, and CISH, were upregulated in P1 (Fig. 4 B). Compared with HCs, P1-derived TCR γδ cells showed significantly increased expression of genes involved in T cell survival (NT5E, COL6A1) (30, 31), expansion (CDKN1C) (32), migration (TIAM1) (33), suppression of T cell response (COL6A1) (34), and inhibition of the JAK-STAT pathway (CDCP1) (35). In contrast, TCR γδ cells from P1 had reduced expression of genes that coordinate inhibition of cell proliferation and migration (SH3BGRL2) (36), T cell activation (ENPP1, IKZF2, CTL4, HK2, IRF4, TNFSF10) (37, 38, 39, 40, 41, 42), and exhaustion (ENPP1, IRF4) (37, 43). Notably, a published transcriptomic signature of TCR γδ T-LGL leukemia (44) was assessed in P1 vs. HCs samples. The top dysregulated genes from this published γδ T-LGL leukemia signature did not cluster P1 vs. HC samples (Fig. S3, D and E).

On the one hand, these transcriptomic data indicate that STAT5-mutated TCR γδ cells are more prone to proliferate. On the other hand, analysis of transcription factor genes (45) in P1’s TCR γδ cells showed decreased expression of genes involved in T cell activation (EGFR2) (46), exhaustion (MAF) (47), memory cell generation (LEF1) (48), and the induction of the Th17 program (RORC) (49) (Fig. 4 C). Additionally, upregulated NR4A3 expression might reduce cytokine production during the effector stage (50) (Fig. 4 C). Next, we focused the analysis on the functional differentiation programs of TCR γδ cells by comparing the transcription of genes associated with the Th1/IFN-γ, Th2, and Th17 effector functions (51) (Fig. 4, D and F). The analysis revealed minimal differences in the Th1/IFN-γ and Th2 gene sets (Fig. 4, D and E). In P1, there were a decrease in the expression of the Th1 cell activation marker gene PDCD1 (Fig. 4 D) and an increase in the expression of the Th2-related gene NOD2, which plays a role in regulating T cell activation (52) (Fig. 4 E).

Interestingly, major differences were observed in genes associated with Th17 differentiation. TCR γδ cells derived from P1 showed higher transcripts for inhibitory NK markers (KIR3DL3, KIR3DL1, KIR2DL4, KIR2DL1) (53), and reduced expression of Th17-associated genes (IL23R, RORC, CCR6, MAF) (54) (Fig. 4 F).

Next, we characterized genes known to be involved in Th17 cell function to further evaluate the differentiation signature of the TCR γδ cells from P1 (51) (Fig. 5 A).

This analysis revealed a decline in several key transcription factors characteristic of Th17 cells (RORC, MAF, LEF1) and other genes associated with Th17 cells (IL17RE, KIT, IL23R, CCR6, IL12RB1, ADAM12) (Fig. 5, A and B). These genes were significantly downregulated in TCR γδ cells from P1 compared with HCs in both unstimulated (Fig. 5 C) and IL-2–stimulated (Fig. S4 A) TCR γδ cells. The repression pattern of these genes in P1 was similar in ex vivo TCR γδ cells (Fig. S4 B). Moreover, the negative regulators of STAT5 hyperactivity, SOCS1 and SOCS2, were elevated in P1, suggesting a compensatory response in STAT5 GOF mutant TCR γδ cells (Fig. 5 A).

The analysis of genes differentially expressed in TCR γδ cells derived from P1 vs. HCs showed an increased expression of IL-9R in P1 (Fig. 5, A and C).

Next, we examined the profile of intracellular cytokines produced by TCR Vδ2 cells from P1 compared with HCs. We found no differences in the production of Th1 cytokines (IL-2, IFN-γ, TNF-α, GM-CSF), Th2 cytokines (IL-4, IL-10, IL-13), Th9 cytokine (IL-9), and Th17 cytokines (IL-17A, IL-22) upon HMBPP stimulation. However, significant differences emerged between the lines after PMA/ionomycin exposure. P1-derived TCR Vδ2 cells showed higher intracellular levels of all measured Th1 cytokines (IL-2, IFN-γ, TNF-α, GM-CSF) compared with HCs. Additionally, Th2 cytokines IL-4 and IL-13 were slightly increased in P1-derived TCR Vδ2 cells (Fig. S4, C and D). Even though the overall Th9 transcriptional profile (55) was not strongly upregulated in P1 compared with HCs (Fig. 5 D), we detected increased IL-9 (Th9) production upon PMA/ionomycin exposure. The cytokines IL-10, IL-17A, and IL-22 were not detected in any of the tested stimulation conditions (Fig. 5, E and F; and Fig. S4, C and D).

JAK inhibitors have been reported to successfully treat autoimmune manifestations associated with germline GOF mutations in STAT3 or somatic GOF mutations in STAT5 (56, 57, 58). To our knowledge, JAK inhibition in patients with TCR γδ cell–specific STAT5 GOF has not yet been documented. We started therapy by administering 4 mg/day of the JAK inhibitor baricitinib under careful clinical, laboratory, and molecular monitoring. Treatment of P1 with the JAK inhibitor markedly reduced the severity and frequency of oral aphthous lesions. It normalized the patient’s reported susceptibility to airway infections and the numbers of monocytes, eosinophils, and neutrophils in the blood (Fig. 6 A). Furthermore, we observed reductions in serum IgE and IgM levels (Fig. 6 B). Initially, the percentages of CD21^low B cells decreased, but did not return to normal reference levels (Fig. 6 A). Hemoglobin levels (Fig. 6 A) and serum IgG and IgA concentrations (Fig. 6 B) remained normal.

Blocking the JAK–STAT signaling pathway with baricitinib did not immediately (within the first 6 mo) alter the imbalance of circulating TCR αβ vs. TCR γδ cells in P1 (Fig. 6 C). The baricitinib treatment did not affect STAT5 phosphorylation in P1-derived TCR γδ cells when exposed to IL-2 in vitro (Fig. 6 D). However, TCR αβ cells from P1 collected during treatment showed lower percentages of pSTAT5 after 4 h of in vitro IL-2 stimulation (Fig. 6 D). There were no differences in the phenotype of TCR Vδ2 cells before and after baricitinib treatment. The marker clusters, including CD95, CD28, CD26, CD56, KLRG1, NKp80, TIGIT, CD69, PD-1, and CD57, were very similar (Fig. 6 E). Additionally, the cluster frequency variation was low before and after treatment (Fig. S5 A). However, a relative decrease in total TCR γδ and TCR Vδ2 cells, associated with a relative increase in TCR αβ cells, was observed over the years under baricitinib treatment (Table 1; and Fig. 6, F and G), with levels currently approaching those seen in healthy donors (Table 1, Fig. 1 D, and Fig. S5 B). While we have not performed NGS-based STAT5B sequencing of sorted TCR γδ cells from P1 in the current sample under baricitinib treatment, Sanger sequencing of TCR γδ cell–derived cDNA as a semiquantitative readout revealed the STAT5 GOF mutation to be near the detection limit (Fig. 6 H). NGS-based TCRγ sequencing of a current (2024) sample under baricitinib treatment reveals persistence of the top clone detected in 2017, while the second most abundant clone from 2017 is now ranked fourth. The top 3 clone from 2017 is now ranked fifth (Table 3). Thus, while the relative (and absolute) TCR γδ cells normalized under baricitinib, the spectrum of TCR clones remains stable over time.

Persistent STAT5 activation caused by somatic mutations has been described to favor leukemogenesis and the growth of several tumors (14, 59). Somatic STAT5 GOF mutations in T cells plus myeloid cells have been described to drive early-onset eosinophilia, urticaria, and dermatitis, T cell lymphocytosis, and lymphoproliferation (60, 61). Additionally, STAT5 GOF mutations are found in myeloid neoplasia in elderly patients, often in conjunction with other somatic mutations (62). GOF STAT5B mutations may drive both primary- or lymphocyte-variant hypereosinophilia (63). Here, we present a detailed, long-term (16-year follow-up) study of a chronic immune disorder driven by a TCR γδ cell–specific STAT5B p.Y665F GOF mutation, associated with a persistent, stable increase in peripheral TCR γδ cells.

How does the observed polyclonal γδ T cell expansion in P1 relate to T-LGL leukemia? T-LGL expansions have a broad spectrum, with T cell clonality of uncertain significance (T-CUS) representing the physiologic/indolent side, and aggressive T-LGL leukemia representing the life-threatening counterpart (24). T-LGL leukemia is classically associated with peripheral lymphocytosis (>2 G/L) caused by clonal expansion of CD8⁺ TCR αβ cells, which often carry STAT3 GOF mutations (64). Notably, classical T-LGL leukemia usually affects older adults, whereas our index patient and other reported pediatric cases of T-LGL expansions (65) were identified at a young age. TCR γδ T-LGL leukemia accounts for only 15% of all T-LGL leukemia cases (66), and therein represents a subset of more severe disease with lower survival rates (67). T-LGL leukemia with STAT5B mutations have been identified in only 2% of TCR αβ T-LGL leukemia patients, the p.N642H mutation being associated with faster leukemic progression (26). One rare form of T-LGL leukemia, characterized by clonal expansion of CD4⁺ TCR αβ cells, is typically associated with STAT5 GOF mutations (68).

T cell clonality is a hallmark of T-LGL leukemia; however, normal adaptive immune function also involves clonal T cell expansion in response to antigens. Thus, interpretation of T cell clonality is not trivial. A recent study analyzing NGS-based human TCR γδ sequencing across the lifespan and tissues demonstrates clonal expansions in adults as the rule rather than the exception (69). Flow cytometry analysis allows only the determination of the TCR Vγ and Vδ chains but is not suitable to dissect the clonality of TCR γδ expansions. The patients analyzed here had normal absolute and relative total T cell counts, an elevated proportion being TCR γδ–positive and CD4-negative. γδ TCR sequencing revealed top clone frequencies between 5 and 15% with <10-fold increase compared with the background, arguing against conspicuous clonality (70). Thus, while there is no black/white distinction between T-CUS and indolent T-LGL leukemia, our patients would be classified within the former. In keeping, unlike typical T-LGL leukemia patients, P1 and all other patients studied here did not display absolute lymphocytosis, anemia, or neutropenia, which are red flags for a hematology referral. This indicates that many patients with similar molecular pathogenesis may remain undiagnosed or receive alternative diagnoses, such as unspecified connective tissue disease or chronic fatigue syndrome. P1, in a remarkable 16-year follow-up, has not yet experienced the usual complications associated with T-LGL leukemia, such as overt autoimmune conditions like rheumatoid arthritis or pure red cell aplasia. However, P1’s quality of life was significantly reduced due to recurring, though usually mild, respiratory tract infections, extensive erosive stomatitis, joint pain in the large joints (shoulders and knees), and chronic fatigue. His systemic hyperinflammatory condition was characterized by neutrophilia (in the absence of smoking), monocytosis, mild eosinophilia, and increased CD21^low B cells, which have recently been shown to expand in response to IFN-γ (71). Notably, the administration of JAK inhibitors is currently uncommon in T-LGL leukemia. Indeed, none of the 137 TCR γδ T-LGL leukemia cases in a recent multicenter analysis received this treatment (67). Instead, methotrexate, cyclophosphamide, and cyclosporine are the main medications used for patients with T-LGL leukemia who have cytopenia or autoimmune conditions. Among the 137 patients with TCR γδ T-LGL leukemia, cyclosporine appeared to be more effective than methotrexate in those with cytopenias (67). In contrast, JAK inhibition has been successfully used in two pediatric patients with severe hypereosinophilia caused by somatic STAT5B GOF mutations in both myeloid and lymphoid lineages (56). In one of these patients, JAK inhibition was complicated by Mycobacterium avium osteomyelitis (72). Very recently, Schmitt et al. described a pediatric case of mosaic p.T628S STAT5B GOF with central nervous system (CNS) demyelinating inflammation, eosinophilia, and lymphocytosis treated with the JAK inhibitor tofacitinib, which was associated with clinical improvement (73). In contrast to P1, the STAT5B GOF mutation was found in ∼50% of reads in all tested myeloid and lymphocyte subsets. Immunophenotypically, mildly increased IgE and IgM were shared with P1, while lymphocytosis and eosinophilia were much more prominent in comparison with P1. P1 also had less acute and systemic disease, as he lacked CNS inflammation and lymphoproliferation, which ultimately required a bone marrow transplant in the p.T628S STAT5B GOF case (73). In P1, JAK inhibition improved reported susceptibility to infection, erosive stomatitis, and most inflammation-related blood biomarkers. However, chronic fatigue and arthralgia remained essentially unchanged. Our findings showed a slow but steady decrease in TCR γδ cells under JAK inhibition, aligning with the expected reduction in STAT5-driven hyperproliferation. We have not yet observed any severe or opportunistic infections in P1 under JAK inhibition. The patient had received extensive vaccinations in accordance with national guidelines for immunosuppressed patients. Besides JAK inhibition, also targeted mTOR inhibition has also been successfully used to control lymphoproliferation in patients with somatic STAT5B mutations (61).

Our understanding of the molecular biology underlying T-LGL expansions remains limited. We could not find any studies that have explored human TCR γδ T-LGL cell activation in response to pAgs. A recent study performed WES and analyzed transcriptomic data from bulk PBMC-derived DNA and RNA in 105 patients with T-LGL leukemia, including 12 with TCR γδ T-LGL leukemia. The study focused on comparing STAT3-mutated vs. STAT3-wild-type T-LGL leukemia samples (74). That study revealed upregulation of interferon signaling pathways, supporting the use of JAK inhibitors in T-LGL leukemia. Only one previous study has characterized transcriptomes in TCR γδ T-LGL leukemia vs. nonleukemic TCR γδ subsets (44), without examining the presence or absence of somatic STAT3 or STAT5 mutations. The study demonstrated that the expression profiles of TCR γδ T-LGL leukemia most closely resembled those of the TEMRA TCR γδ subset in HCs. The published transcriptomic T-LGL leukemia profile did not match that of P1’s TCR γδ cells. The IFNG gene, which encodes IFN-γ, was highly upregulated, and the expression of pro-apoptotic genes was lower in TCR γδ T-LGL leukemia cells than in HCs (44).

We have examined the biological implications of STAT5 GOF in TCR Vδ2 cells using three approaches. First, flow cytometry showed that P1-derived TCR Vδ2 cells expressed a more innate-like phenotype and had enhanced effector functions compared with Vδ2 cells from HCs. This was evident from the increased expression of CD26, CD56, and NKp80 in P1 cells (75, 76, 77).

Second, we studied how TCR Vδ2 cells respond to the prototypic pAg HMBPP. T cells exposed to HMBPP exhibited stronger TCR-mediated proliferation of TCR Vδ2 cells from P1 vs. HCs. Also, ex vivo analysis of these cells showed increased CD57 expression in HCs compared with P1. In CD4⁺ and CD8⁺ T cells, CD57-positive cells are more differentiated (27) and have impaired proliferative capacity (27, 78, 79). This is consistent with the higher proliferative potential of P1 cells, which express lower levels of CD57 than HCs. The differences in proliferation were observed under suboptimal T cell stimulation, whereas cells from healthy donors can proliferate robustly when challenged with a strong stimulus, such as PHA. Both observations support that the STAT5 GOF mutation is linked to enhanced cell proliferation, in keeping with the relative expansion of TCR Vδ2 cells in P1’s peripheral blood. The gene transcription analysis of P1’s TCR γδ cells consistently revealed higher expression of genes associated with survival and expansion (NT5E, COL6A1, CDKN1C) (30, 31, 32). Genes that control cell growth were also found to be overexpressed in P1 (COL6A1, CDCP1, SH3BGRL2, SOCS1, and SOCS2) (34, 35, 36, 80). This suggests an intact negative feedback mechanism counteracting the STAT5 GOF–dependent persistent T cell expansion, eventually precluding progression to T-LGL leukemia.

At the transcriptional level, TCR γδ cells from P1 showed changes in IL-2 signaling, as well as in genes related to Th17 and Th9 differentiation. STAT5B GOF TCR γδ cells showed decreased expression of Th17 signature genes, including RORC, MAF, LEF1, IL17RE, KIT, IL23R, CCR6, IL12RB1, and ADAM12. This resembles the transcriptional changes observed in TCR γδ cells from c-MAF knockout mice, which show a block in IL-17 differentiation during early thymocyte development (81). Lower expression of the MAF, RORC, and CCR6 genes might result from ongoing STAT5 activity. STAT5 directly targets and suppresses transcription of these genes (82, 83, 84, 85, 86), thus blocking Th17 differentiation.

Activation of TCR Vδ2 cells with HMBPP did not show any differences in the levels of Th1-, Th2-, Th9-, or Th17-related cytokines. However, stimulation with PMA/ionomycin led to a higher overall accumulation of Th1 cytokines, accompanied by a subtle accumulation of Th2 cytokines and IL-9 in TCR Vδ2 cells from P1 compared with HCs, consistent with the changes observed at the transcriptional level. No Th17 cytokines were detected in this set of experiments.

The unstimulated or IL-2–stimulated condition from the RNA-seq experiment does not necessarily reveal the full cytokine potential of the T cells (87, 88). In contrast, PMA/ionomycin stimulation used in the T cell lines induces a supramaximal signaling that exposes the potential of cytokine production of T cells, showing differences that might not be visible in untreated or IL-2–treated conditions (89). Recently, a model has been proposed for Th17 commitment in TCR γδ cells. This model suggests that the transcription factor SOX13 activates c-MAF, which in turn activates RORγ (51). Based on this model and the observed transcriptome of P1, the STAT5 GOF might suppress MAF and RORC expression, as seen in TCR γδ cells from c-MAF knockout mice (81). Notably, STAT5 plays a crucial role in Th2 differentiation. Although it does not enhance the expression of the Th2-related gene GATA3, it does activate the expression of the IL-4 gene (90). We observed higher levels of both intracellular and secreted IL-4 and IL-13 in TCR Vδ2 cell lines from P1 patients than in those from HCs.

Additionally, TCR Vδ2 cells from P1 patients exhibited higher intracellular IL-9 levels than those from HCs. Overall, the transcriptomic analysis revealed STAT5-related changes in gene expression, including suppression of Th17 signature genes, and imbalances among the Th1, Th2, and Th9 cytokine signatures. A limitation of our study is that we analyzed bulk TCR γδ cells and thus cannot functionally discriminate between STAT5B-mutated vs. STAT5B-wild-type cells at the single-cell level.

It remains unclear whether an underlying cause triggered the acquisition and/or maintenance of STAT5B GOF. Chronic viral infections caused by CMV, EBV, and HIV are suspected to contribute to the context of T-LGL leukemia (91, 92). However, there was neither clinical nor laboratory evidence of a chronic viral infection, and the EBV PCR in peripheral blood was consistently negative, even under JAK inhibition. WES in P1 identified two additional germline mutations in TYK2, one with a CADD score of 33 in a highly conserved region, together with another synonymous mutation (Fig. 1 D and Table 4). TYK2 loss-of-function resulting from compound-heterozygous or homozygous mutations has a distinct unrelated clinical phenotype (93). Germline-activating TYK2 mutations have been identified as a predisposing factor to the development of acute lymphoblastic leukemia in children (94). Therefore, it is possible that dysregulated TYK2 activity contributed to the development and/or maintenance of STAT5B GOF–mutated TCR γδ cells, experimental analysis of which is beyond the scope of this manuscript. A very recent comprehensive study of T-LGL leukemia patients has revealed an unexpectedly high co-occurrence of rare, damaging germline variants in genes linked to IEI (95).

In summary, a patient with a chronic hyperinflammatory immune dysregulation carried a TCR γδ cell–specific STAT5B GOF mutation despite normal peripheral lymphocyte counts. Our search within our own multicenter prospective cohort of patients with immune dysregulation reveals several other patients with a similar clinical and immunophenotypic profile. In all those patients, TCR sequencing revealed polyclonal T cell populations. The culprit immunophenotypic anomaly (i.e., an altered TCR αβ vs. TCR γδ ratio) in such patients may remain unnoticed. We advise that individuals with chronic immune dysregulation, normal absolute lymphocyte counts and a higher proportion of CD4⁻CD8⁻ T cells should be analyzed for relative TCR γδ cell expansion. Among such patients, we identify an individual with a TCR γδ cell–specific STAT5B GOF mutation, suitable for targeted treatment with JAK inhibitors. We have not yet performed a thorough molecular dissection of all other individuals identified with chronically altered TCR αβ vs. TCR γδ ratios listed in Table 1. In one patient (cohort number 86), WES analysis of gDNA from sorted TCR γδ cells detected neither STAT3 nor STAT5 GOF mutations that could have explained the expansion of TCR γδ cells. However, this individual’s TCR Vδ2 cells, comprising the majority of his TCR γδ cells, showed high reactivity to MR1, an MHC class I–like molecule, suggesting an antigen-driven expansion of these cells (96). Besides STAT5-dependent and antigen-driven expansion, TCR γδ cell generation, maintenance, and function are affected by numerous other IEI genes and associated signaling pathways, as recently summarized by Sagar and Ehl (97).

Within a cohort of patients with chronic susceptibility to infection, a subset exhibits a persistently imbalanced TCR αβ vs. TCR γδ ratio as the primary phenotypic anomaly, which may be easily overlooked in standard immunologic assessments. Underlying molecular reasons in such patients include TCR γδ–specific STAT5B GOF, which may be addressed by tailored immune modulation.

The cohort and associated immune cell characterization have been approved by the Ethical Commission of Northwestern and Central Switzerland (EKNZ-2015-187).

An experienced clinical immunologist clinically evaluates patients, and only those with likely primary immune dysregulation (i.e., excluding patients with secondary immune dysregulation) are enrolled.

There is written consent for the use of the patient-derived CT-based imaging and images.

Sex and age were taken as biological variables. HCs and P1 were sex- and age-matched for sample collection.

PBMCs and granulocytes were isolated using standard density-gradient centrifugation (Lymphoprep, Axis-Shield). To isolate granulocytes, the erythrocyte/granulocyte fraction was lysed with RBC lysis buffer (#158103; Qiagen). Monocytes and eosinophils were then isolated using a negative selection kit from Stemcell, following the manufacturer’s instructions.

THP-1 is a human monocytic leukemia cell line obtained from the American Type Culture Collection (No. TIB-202). The tumor cells were grown in RPMI 1640 medium with 10% heat-inactivated fetal calf serum (FCS), 1 mM sodium pyruvate, 1× nonessential amino acids, 1× stable glutamine, and 50 µg/ml kanamycin at 37°C in 5% CO₂. All reagents were purchased from BioConcept. We regularly checked that the cell lines were free of Mycoplasma contamination.

TCR Vγ9Vδ2 cell lines were sorted by flow cytometry from PBMCs using anti-Vδ2 and anti-Vγ9 mAbs. T cell lines were then cultured and expanded in the presence of 100 U/ml of human recombinant IL-2 (#200-02; PeproTech), 1 µg/ml of PHA (#30852801 HA16; Remel), and irradiated PBMCs (5 × 10⁵ cells/ml).

For proliferation assays, PBMCs were labeled with carboxyfluorescein succinimidyl ester (#C34554; Thermo Fisher Scientific) or CellTrace Violet (#C34557; Thermo Fisher Scientific) according to the manufacturer’s instructions. The labeled cells were then stimulated with 0.4 nM HMBPP (#95098; Sigma-Aldrich) or 1 µg/ml PHA (#R30852701, Thermo Fisher Scientific). Proliferation and the upregulation of CD69, CD137, CD39, and ICOS were assessed by flow cytometry on day 3.5 after stimulation.

For the activation assay, THP-1 cells were preincubated with different doses of HMBPP for 3 h at 37°C, then cocultured with TCR Vγ9Vδ2 cell lines from either HCs or P1 for 18 h at 37°C and 5% CO₂, at an effector-to-target ratio of 1:2. Activation was measured using ELISA.

Release of IFN-γ, TNF-α, IL-4, and IL-13 was measured using ELISA, as previously described (98). The methods used were human IFN-γ (capture MD-1 mAb [#507502]; revealing biotinylated 4S.B3 mAb [#502504]; BioLegend); human TNF-α (capture Mab1 mAb [#502802]; revealing biotinylated Mab11 mAb [#502904]; BioLegend); human IL-4 (capture 8D4-8 mAb [#500702]; revealing biotinylated MP4-25D2 mAb [#biotinylated MP4-25D2]; BioLegend); and human IL-13 (capture JES10-5A2 rat Ab [#10125-01]; revealing biotinylated SB126d Rat Ab [#15930-08]; SouthernBiotech).

Clinical routine immunophenotyping was performed as previously described (99) (see Table 2). Clinical routine hemato-immunologic phenotyping, including T-LGL–related markers, was performed as previously described (100) (see Table 1).

Antibodies used for research-based flow cytometry, including clones, are listed in Table 5.

Cells were stained with Zombie NIR (#423106; BioLegend) or LIVE/DEAD BLUE (#L23105; Thermo Fisher Scientific) fixable viability dyes in PBS for 20 min at 4°C, then surface-stained. Anti-human mAbs were used to stain cells in FACS buffer (PBS, 0.5% BSA, 0.02% NaN3) at 4°C for 20 min. Viability was assessed by adding DAPI to PBS after cell surface staining. Cells were acquired using an Aurora spectral analyzer (Cytek) or a CytoFLEX flow cytometer (Beckman Coulter) and analyzed with FlowJo v10 software (LLC). Cell sorting was performed by flow cytometry on a BD Influx, a BD Aria III, or a FACSMelody Cell Sorter (BD) cytometer.

Staining of pSTAT5 (Y694 STAT5A/Y699 STAT5B) was done on sorted TCR γδ and TCR αβ cells that had been rested overnight in medium at 37°C. The medium used, called complete medium, was RPMI (Sigma-Aldrich), supplemented with 100 U/ml of penicillin–streptomycin (Gibco), 1× MEM nonessential amino acid solution (Gibco), 1× GlutaMAX (Gibco), and 10% FCS (Gibco). The next day, 2 × 10⁵ cells were either stimulated with hIL-2 (300 U/ml, Novartis) for 30 min or left unstimulated at a concentration of 2 × 10⁵ cells/ml. The pSTAT5 antibody from BD Pharmingen, used for flow cytometry, was recently used to characterize STAT5B GOF patients (60). Staining of intracellular cytokines was performed in TCR Vδ2 cell lines treated with 2 µM monensin (#420701; BioLegend) and 20 µg/ml brefeldin A (#420601; BioLegend). The cells were then either stimulated or not with 50 ng/ml PMA (Sigma-Aldrich) and 500 ng/ml ionomycin (Sigma-Aldrich).

Cells were fixed using Fix/Perm buffer (#554714; BD Biosciences) in the dark at room temperature for 20 min. Next, cells were washed in PBS with 2% FCS and then permeabilized with Perm III Buffer (#558050; BD Biosciences) at 4°C for 20 min.

For STAT5 staining, the conjugated antibody and isotype control were incubated at room temperature for 45 min, followed by a 45-min incubation with the secondary antibody at the same temperature.

For intracellular cytokine staining, cells were incubated with cytokine-specific antibodies for 30 min at 4°C. We collected cells using either a CytoFLEX (Beckman Coulter) or an Aurora spectral analyzer (Cytek). Data were analyzed with FlowJo v10 software (FlowJo LLC).

Endpoint PCR was performed using genomic DNA or cDNA with GoTaq Polymerase (#M7841; Promega) following the manufacturer’s instructions. The primer concentration was 0.5 µM. For cloning the PCR product, Phusion Polymerase (#M0530S; NEB) and the Zero Blunt TOPO PCR cloning kit (#K287540; Thermo Fisher Scientific) were used.

PCR was performed on T Professional TRIO PCR Thermocycler (Core Life Sciences). The PCR products were separated and visualized using a 1.5% agarose gel. Specific bands were then excised and purified using the Qiagen Gel Extraction kit (#28704; Qiagen). Sanger sequencing was conducted by Microsynth.

Primers for the STAT5B Y665F mutation on the cDNA level:

Fw: 5′-AGG​GAG​AAT​TTA​CCA​GGA​CGG-3′

Rev: 5′-TCC​ATG​TAC​GTG​GCG​CTG-3′.

Primers for STAT5B Y665F mutation on the genomic DNA level:

Fw: 5′-AAA​TGG​AGA​TTT​CTA​TTG​GAG​CCA​T-3′

Rev: 5′-TAG​CAG​ACT​CGC​AGG​GAA​CT-3′.

WES was performed using genomic DNA from CD3⁺TCR γδ⁺ and CD3⁺TCR αβ⁺ cells sorted by flow cytometry ex vivo.

WES has been performed as recently described (101). The presence of rare mutations linked to immunodeficiencies was investigated using a compiled list of primary immunodeficiencies (102). The genome viewer confirmed all reported variants. Immunodeficiency-related gene variants were listed when the CADD score was >15, and the minor allele frequency listed in gnomAD (https://gnomad.broadinstitute.org) was <0.01.

Blood or PBMC-derived DNA was used for amplification for TCRγ- or TCRβ-specific CDR3 regions (75–100 base pairs) using Oncomine TCR Pan-Clonality Assay. Data analysis was performed using Torrent Suite and Ion Reporter software.

Cells were lysed in TRIzol reagent (#15596026; Thermo Fisher Scientific), and then, RNA was purified using the Direct-zol RNA kit (#R2061; Zymo Research). The RNA concentration was measured with the Qubit fluorometer (Thermo Fisher Scientific). Library preparation was done at the University of Basel by the Genomic Core Facility. Sequencing was performed on an Illumina NovaSeq S1, yielding 48–60 million reads per sample.

Reads were aligned to the human genome (UCSC version hg38 analysis set) using STAR (version 2.7.0c) with additional options “outFilterMultimapNmax10-- outSAMmultNmax 1.” This allowed multimapping reads to be included, with a maximum of 10 locations and a random selection of one location. The output was then merged across flow cells, sorted, and indexed using SAMtools (version 1.9). We evaluated read and alignment quality using the qQCReport function from the QuasR Bioconductor package (R version 3.5.0, Bioconductor version 3.8). To count the number of reads (5′ ends) overlapping with each gene’s exons, we used the featureCounts function from the subread R/Bioconductor package, assuming an exon union model. Gene annotation was based on Ensembl version 94.

Cells were lysed using TRIzol reagent (#15596026; Thermo Fisher Scientific). We then isolated RNA using chloroform (#319988; Sigma-Aldrich), followed by 15-min centrifugation at 4°C and 14,000 g, according to the manufacturer’s instructions, and purified it with QIAamp RNA Blood Mini Kit (#52304; Qiagen). To determine the RNA concentration, we used the NanoDrop 2000c (Thermo Fisher Scientific). For DNA digestion, we used the DNase I Amplification–Grade Kit (#18068015; Sigma-Aldrich). Concentrations of 100 ng/μl to 1 μg/μl RNA were used for cDNA synthesis. We annealed random primers (#C1181; Promega) at 70°C for 5 min. cDNA synthesis was done according to the Qiagen GoScript Reverse Transcription System protocol in T Professional TRIO PCR Thermocycler (Core Life Sciences).

Flow cytometry data were exported from FlowJo v10.7.1 and imported into R v4.2.0 using the Bioconductor package flowCore v2.10.0 (103) to perform clustering analysis. Marker expression levels were transformed using the inverse hyperbolic sine transformation (asinh function in R) with a cofactor of 150. Clusters were then computed using R’s implementation of PhenoGraph (104), based on the transformed expression values for nine markers (CD69, CD26, CD28, CD56, CD57, CD161, NKp80, CD95, and PD-1) and a k value of 100. A heatmap showing the average expression of each marker in each cluster was produced using pheatmap v1.0.12 (105), and the clusters were visualized as colors overlaid on a Uniform Manifold Approximation and Projection (UMAP) using the R implementation of UMAP (v0.2.10.0) (106, Preprint).

Statistical analysis and data visualization were performed using GraphPad Prism (GraphPad Software, Inc.). P values are indicated in the figure panels and legends.

The assembling of figure panels was done using Affinity Designer 2.

The following supplemental materials are provided for this manuscript. Fig. S1 depicts immune cell populations and imaging of P1, as well as STAT5B sequencing in enriched myeloid cell populations. Fig. S2 characterizes the immunophenotype of TCR Vδ2–negative cells from P1. Fig. S3 illustrates proliferation and activation assays of TCR Vδ2 cells challenged with HMBPP in P1 vs. HCs. Fig. S4 compares gene expression in TCR γδ cells and cytokine profiling in P1 vs. HCs. Fig. S5 characterizes P1-derived TCR Vδ2 cells before and under baricitinib treatment.

The datasets used in this study are not publicly available to protect participant/patient anonymity. Requests to access the datasets can be made to the corresponding author. The disease-associated somatic c.A1994T STAT5B mutation has been submitted to ClinVar (accession number: SCV007540363).

We thank Andrew Chancellor for the preliminary flow cytometry studies. Calculations were performed at the University of Basel’s sciCORE (http://scicore.unibas.ch/) scientific computing center.

M. Recher was supported by the Swiss National Science Foundation (SNSF) (Grant Number: 310030_192652), the Novartis Foundation for Medical–Biological Research (Grant Number: 24A060), and the Nora van der Meeuwen-Häfliger Foundation. G. De Libero was supported by grants from the SNSF (310030-173240 and 310030B-192828), the Swiss Cancer Research Foundation (KFS-4707-02-2019), and the Basel Cancer League (KLbB-4779-02-2019).

Author contributions: Benedikt J. Meyer: conceptualization, data curation, formal analysis, investigation, methodology, project administration, software, supervision, validation, visualization, and writing—original draft, review, and editing. José Pedro Loureiro: conceptualization, data curation, formal analysis, investigation, methodology, project administration, validation, visualization, and writing—original draft, review, and editing. Vladimir Nosi: formal analysis. Robin Hupfer: investigation. Adhideb Ghosh: formal analysis and software. Fabio Poletti: data curation, formal analysis, investigation, and software. Annaïse Jauch: formal analysis. Julia Hirsiger: data curation. Caroline Berkemeier: formal analysis and investigation. Ingmar Heijnen: investigation. Jan Dirks: investigation and methodology. Ilaria Alborelli: formal analysis. Thomas Menter: investigation, methodology, validation, and writing—review and editing. Alexandar Tzankov: data curation, funding acquisition, investigation, methodology, supervision, validation, and writing—review and editing. Christoph Hess: conceptualization, methodology, and writing—original draft, review, and editing. Alexander A. Navarini: conceptualization, funding acquisition, and writing—review and editing. Christoph T. Berger: investigation and writing—review and editing. Lucia Mori: funding acquisition and writing—review and editing. Gennaro De Libero: conceptualization, data curation, funding acquisition, project administration, resources, supervision, and writing—original draft, review, and editing. Mike Recher: conceptualization, data curation, funding acquisition, investigation, methodology, project administration, resources, supervision, validation, and writing—original draft, review, and editing.